Parasite clones and human foreskin fibroblasts (HFF) were grown and maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum (Hyclone), penicillin and streptomycin (50 μg/ml each) and 200 μM L-glutamine. We will refer to this complete medium as D10. T. gondii were maintained by serial passage in HFF monolayers at 37°C in a humid 5% CO₂ atmosphere, including RH parasites (National Institutes of Health AIDS Reference and reagent Repository, Bethesda, MD), RH tachyzoites expressing cytoplasmic yellow fluorescent protein (YFP) (kindly provided by M.J. Gubbels, Boston College, Boston, Massachusetts) (Gubbels et al., 2003) and Ku80 knock-out parasites (ΔKu80, kindly provided by V. Carruthers, University of Michigan Medical School Dexter, Michigan) (Huynh and Carruthers, 2009; Fox et al., 2009).

We mutagenized T. gondii with ENU as previously described (Pfefferkorn and Pfefferkorn, 1979; Nagamune et al., 2007). Briefly, approximately 1.5 × 10⁷ intracellular tachyzoites of wild-type RH strain T. gondii were mutagenized with ENU (N-Ethyl-N-nitrosourea, N3385-1G, Sigma Aldrich; 100–200 µg/ml) in DMEM without FBS for 2 h at 37°C. These parasites were washed and transferred to a new confluent HFF monolayer for selection with auranofin (2.5 μM-3 μM) until normal pace of replication was observed (~2 weeks). Auranofin resistant clones were single cell cloned by limiting dilution in 2 μM auranofin. Isolated clones were propagated in D10 media with 1 μM auranofin to maintain selection pressure without cytotoxic effects on host cells (data not shown).

We selected ten representative T. gondii clones from independent ENU-generated pools to investigate resistance. To assess auranofin resistance, we modified a tachyzoite growth competition assay (Ma et al., 2008). Using wild-type YFP-expressing T. gondii parasites as a control, confluent monolayers of HFF cells were inoculated with approximately equal numbers (2 × 10⁶; 1:1 ratio) of ENU-mutagenized parasites and YFP- expressing wild-type RH parasites. Upon lysis, a new T25 flask with confluent HFF was inoculated with 0.25 ml of lysed parasites. Another 0.25 ml of lysed parasites was used for flow cytometry. The YFP fluorescence allowed us to differentiate wild-type and Aur^R parasites in the mixed populations to track their growth over serial passages. Host cell debris was removed by filtering of parasites with a 3 μm filter membrane. After parasites were collected by centrifugation (400g for 10 min), the pellet was resuspended in 1 ml PBS. The total number of parasites and the number of parasites expressing YFP were quantified by flow cytometry using a FACSCalibur flow cytometer (Becton Dickinson) and analyzed with FlowJo software (Tree Star). The result of three independent experiments analyzing the replication fitness advantage between each auranofin resistant clone and the wild-type parasites was analyzed with parametric paired t-tests. A two-tailed p-value <0.05 was considered statistically significant.

We extracted genomic DNA (gDNA) from auranofin resistant parasites lines using the DNeasy Blood and Tissue kit (Qiagen). The thioredoxin reductase gene (TGGT1_309730, ToxoDB.org) was amplified from gDNA samples as two PCR fragments ~3.7 kb each. The 5’ half of the gene was amplified with primers GGACCTATGAGTTGCCTCTCTG and GTCAAATCCCAATTCGCGGAG, and the 3’ half with GTCTGCTCTCTCTGTTTCGCAAG and CCTCCACATCCTCCAGAAGC. Sequencing primers are 5TrxR 1R, 5TrxR 1F and 5TrxR 3F for the 5’ half, and TrxR F 8, 3TrxR 2F, 3TrxR 3F, and 3TrxR 4F for the 3’ half ( Table S1 ).

To determine the presence of other SNVs in the coding region of our T. gondii auranofin resistant clones, gDNA from 10 resistant clones and wild-type RH parasites was submitted for library construction and whole genome sequencing at the Genomic High Throughput Facility at University of California, Irvine. Briefly, genomic DNA (gDNA) was initially quantified with a Qubit DNA High Sensitivity kit. One hundred nanograms of gDNA was sheared to <300–500> bp in size using a Covaris S220 ultrasonicator (shearing conditions: duty Cycle 10%, Cycles/burst 200 and treatment time 55 sec) (Covaris, MA). The size and the quality of the resulting DNA fragments were analyzed with an Agilent DNA High Sensitivity chip. Ten nanograms of each sample’s fragmented DNA was end-repaired, poly adenylated and ligated to a 6bp DNA barcodes (Bioo Scientific NextFlex DNA barcodes) using a PrepX Complete ILMN DNA Library kit (Takara Bio, CA). The genomic library was built using the Apollo 324 system (Takara Bio, CA) to generate 520 bps inserts. To enrich the DNA libraries, 8 cycles of PCR amplification using a Kapa library amplification kit was used (cycling conditions: 98°C for 45 s, 8 cycles of 98°C for 15 sec, 60°C for 30 s 72° C for 30 s and then 72°C for 1 min) in an Apollo 324 system. The ultimate DNA concentration was calibrated using Qubit 2.0 dsDNA HS Assay kit and analyze with an Agilent DNA High Sensitivity chip. Before sequencing, we quantified our DNA libraries using a KAPA Library Quantification kit for Illumina platforms (Roche). Each library was normalized to 2nM and then pooled equimolar amounts in the final multiplex library that was sequenced on the Illumina HiSeq 4000.

The ten auranofin resistant T. gondii clones and wild-type RH T.gondii were sequenced with HiSeq 4000 paired end 100 reads, quality analyzed, adapter and quality trimmed before alignment with the annotated reference genome of GT1 strain from ToxoDB.org (published 7/19/2013) (Gajria et al., 2008). After initial quality control with FASTQC, reads were trimmed using Illumina adapter sequences with the error probability of 0.05 and reads shorter than 20 bases were discarded. Single Nucleotide Variant (SNV) calling for the haploid genome had a coverage threshold of >5 fold. Each of the 10 mutant clones were compared to wild-type RH variant track and shared variants were filtered out. Only SNVs in the coding regions of the 10 mutant clones are reported. Only genes containing SNVs with 100% frequency and ≥50% of read count and read coverage were considered in our final analysis. Genomic DNA from each mutant clone was used to amplify the targeted SNV region of interest. The resulting amplicons ranged between 1.7–4.5 kbp in size. PCR amplicons were purified prior to sequencing via a QiaQuick PCR Purification Kit (Qiagen 28106). Amplicons were sequenced by Sanger sequencing (Genewiz), using sequencing primers targeted to the SNV region ( Table S1 ). Genomic sequences for all 10 mutant lines can be accessed NCBI, under the Bioproject PRJNA679812.

We generated a T. gondii clone with a 3’ knock-in YFP tag to TgSOD2 using a ligation independent cloning approach as previously described (Huynh and Carruthers, 2009). The mitochondrial distribution of our clone was confirmed by fluorescent microscopy using a Zeiss Axioskop with Axiovision camera and software. Co-localization of the TgSOD2-YFP with mitochondria was done by staining parasite mitochondria with an anti-F1B ATPase antibody (gift from Peter Bradley at UCLA). We subsequently introduced the L201P mutation into the TgSOD2-YFP line using a modified CRISPR plasmid generated using plasmid pSAG1::CAS9-U6::sgUPRT as the scaffold (Shen et al., 2017; Addgene # 54467). Our plasmid had an additional DHFR-TM marker for pyrimethamine selection derived from pLoxP-DHFR-mCherry [(Long et al., 2016), Addgene # 70147], and the two gRNA sequences were replaced with cggtataccggacagatccg and TAGTTTCGCCCAGTTCAAGG, selected by the Benchling software (Benchling.com).

The DNA sequence used for homology directed repair (HDR) was amplified with the primers GATAGTGTGTGAAGAGCAGC and CTGCCGTTACCAACATGG from the LIC-YFP-SOD2 plasmid generated above. PCR amplicon contained the L201P mutation (CTA->CCA) generated with Q5 Site-Directed Mutagenesis kit (New England Biolab), as with the following modifications. To screen for positive CRISPR-Cas9 clones containing L201P, two silent mutations were created to generate new and unique restriction endonuclease sites close to Leucine 201: AflI on Leucine 180 (CTC->CTT, 62 bps upstream of L201P) and XcmI on Serine 208 (TCA->TCT, 22 bps downstream of L201P). The PCR amplicon contained two additional deletions corresponding to the gRNA sites on endogenous SOD2 gene, in order to avoid SpCas9 protein cleaving the amplicon for HDR. The sequence AGAGACCCGGCGG (including the PAM sequence cgg) was deleted on the HDR sequence corresponding to gRNA #1, and sequence CACTTAACAGAGGGTC (including the PAM sequence agg) was deleted on the HDR sequence corresponding to gRNA #2.

The L201P point mutation was introduced into the TgSOD2.YFP line through CRISPR-Cas9 ( Figure S1 ). Transfection of the CRISPR-CAS9 plasmid and the HDR DNA was carried out as published previously, with minor modifications (Shen et al., 2017). Briefly, 10⁷ of freshly lysed parasites with SOD2-YFP were filtered and spun down at 400xg and 18⁰C for 10 min. The pellet was resuspended with Cytomix buffer (10 mM KPO₄, 120 mM KCl, 5 mM MgCl₂, 25 mM HEPES, 2 mM EDTA, 2µM ATP and 5µM glutathione) up to 800µl, including 7.5µg of CRISPR plasmid and 1.5µg of HDR DNA. The mixture was transfected using the ECM 630 Electro Cell Manipulator (BTX). Parasites were incubated into T25 flasks after electroporation and switched to 2 μM pyrimethamine selection 24 h later. Transfected pools remained in pyrimethamine selection for 3 days, and then switched to D10 media without selection for one more day before single cell cloning. The gDNA for isolated clones were collected by the DNeasy Blood and Tissue kit (Qiagen) and used to PCR amplify SOD2 with either primers SOD2 F/AmpR (for SOD2-YFP) or primers SOD E2/SOD R (for SOD2 only, Table S1 ). PCR amplicons were digested with XcmI for screening and positive clones were sequence verified.

To expose our parasites to oxidative stress, we harvested freshly lysed parasites and treated 5 × 10⁴ parasites with 500 μM hydrogen peroxide (H₂O₂) for 30 min at 37°C, 5% CO₂, with or without 1 μM auranofin After incubation, ROS was detected using H₂DCDFC (Invitrogen). Each sample was incubated with 1 µM of H₂DCDFC and examined under a Spectramax i5, (Molecular Devices California US; multimode microplate reader SoftMax Pro 7.0.2 Software) multimode microplate reader (excitation 494 nm; emission 522 nm) at 37°C for 15 min. For excitation, a single flash from a UV Xenon lamp was used for each well, and emission signals were recorded with a sensitivity setting of 100. The values are presented as relative fluorescence units. We compared ROS accumulation of each representative auranofin resistant T. gondii clone to that of wild-type parasites by an unpaired t-test in three independent experiments. A two-tailed p-value <0.05 was considered statistically significant.

The predicted amino acid sequences for TgSOD2 (TGGT1_316330) and TgRNR (TGGT1_294640) were submitted to the SwissModel site to identify the optimal available protein structures for modeling (Guex et al., 2009; Benkert et al., 2011; Bertoni et al., 2017; Bienert et al., 2017; Waterhouse et al., 2018). The Toxoplasma sequences were threaded onto the structure of Plasmodium knowlesi SOD (PDB 2AWP) (Vedadi et al., 2007) and human RNR (PDB 3HND) (Fairman et al., 2011). Clustal Omega was used to align the amino acid sequences and EMBOSS was used to calculate percent identity and similarity. Information on MS peptide coverage was obtained from the ToxoDB database (Gajria et al., 2008).

We used qPCR to quantify the expression level of thioredoxin reductase gene in both wild type and mutant parasites. RNA was extracted from parasite lines using the RNeasy Mini Kit (Qiagen), and then used for cDNA synthesis via iScript cDNA Synthesis Kit (Bio-Rad). The resulting cDNA samples were then subjected to qPCR with iTaq Universal SYBR Green Supermix (Bio-Rad) with the following gene targets: actin (TGGT1_209030, ToxoDB.org), GAPDH1 (TGGT1_289690, ToxoDB.org), thioredoxin reductase (TGGT1_309730, ToxoDB.org), and superoxide dismutase 2 (TGGT1_316330, ToxoDB.org).

The actin gene was amplified from cDNA samples as one PCR fragment ~60bp, with primers CACGAGAGAGGATACGGCTTC and CGATGTCGCTAGAGTCCTCAG. The GAPDH1 gene was amplified in one PCR fragment ~120bp with primers CACTTTGCCAAGATGGTGTG and GACTGTCTTCAACGAGAAGGAG. The thioredoxin reductase gene was amplified a ~70bp PCR fragment using the primers CACAAACACGAACAACGGCG and GTGGACGGCTGAACAAAGTCG. For the ~50bp fragment in superoxide dismutase 2 gene, primers GGCACACCGTTCGCTGATAAG and CGTGGTTCCATGCTTGTGCTG were used.

In order to quantify the amount of target transcript present, target CT values were normalized against actin CT values. The resulting ΔCT values were then compared through 2-way ANOVA between the different parasite strains surveyed.

To verify auranofin resistance of the selected clones, we adapted a previously described competition assay (Ma et al., 2008). This was used in place of a normal dose-response assay because at auranofin concentrations higher than 3 μM, the fibroblast monolayer begins to disrupt (data not shown), making the plaque assays not feasible. To confirm resistance to auranofin in the Aur^R T. gondii lines, we assessed their growth in competition with YFP-expressing RH strain parasites in control media and in media with 2.5–3.0 μM auranofin. In competition assays, the inoculating parasite population is composed of 50% YFP-expressing wild type (sensitive) parasites and 50% of an Aur^R line verified by flow cytometry. The relative contribution of the individual populations was measured after each cycle of lytic growth in serial culture. If a parasite line increases to >50% of the population, it exhibits a growth advantage over the competing line. In the absence of auranofin, Aur^R lines grew comparably or less well than the parental RH line that does not express YFP ( Figure 1A ). However, in the presence of 2.5 μM auranofin, all Aur^R lines displayed significant growth advantage over YFP-expressing wild-type RH parasites ( Figure 1B ). The observations indicated that day 6 is the optimal day to capture the growth advantage of resistant lines under auranofin selection because YFP-expressing wild-type parasites were eliminated from the culture. Given these results, we measured the ability of a larger set of Aur^R lines to compete with wild-type YFP-expressing RH strain parasites in the absence or presence of auranofin by quantifying their contribution to the culture at 6 days ( Figure 2 ). These results show that 9 of the 10 Aur^R lines outcompete the sensitive YFP-expressing RH line in the presence of 2.5 μM auranofin. Line 6 has a distinct profile: it grows slowly in both control and auranofin media, suggesting that it may evade auranofin-mediated killing by protracted replication.

Considering that thioredoxin reductase (Sannella et al., 2008; Angelucci et al., 2009; Debnath et al., 2012; Ilari et al., 2012) is the proposed target of auranofin, we hypothesized that resistance in T. gondii would develop through mutations to the parasite thioredoxin reductase gene (TgTrxR, TGGT1_309730). Therefore, we amplified and sequenced the TgTrxR gene in 53 lines derived from parallel (independent) selections (not shown). We did not detect mutations (SNVs) in the thioredoxin reductase gene of any of the lines, indicating that auranofin resistance does not arise through modifications to TgTrxR that directly influence auranofin binding or TgTrxR enzyme activity.

The anti-parasitic activity of auranofin relies on its ability to induce accumulation of ROS in parasites (Debnath et al., 2012). We exposed wild type RH strain parasites and four representative Aur^R lines to hydrogen peroxide (H₂O₂) for 30 min before measuring ROS accumulation with dichloro-fluorescein (H₂DCDFC). All Aur^R lines displayed decreased accumulation of ROS relative to the parental RH line in both the absence and presence of auranofin ( Figure 3 ). For control, we also measured ROS accumulation in the absence of H₂O₂ and observed no fluorescence signal (data not shown). This negative control is consistent with a prior study measuring ROS accumulation using the same dichloro-fluorescein dye without the H₂O₂ (Lee et al., 2006).

Given that we did not observe mutations in T. gondii thioredoxin reductase, we performed whole genome sequencing (WGS) analysis of 10 independent T. gondii Aur^R lines. Our goal was to determine whether these T. gondii parasites harbor a common mutation(s) that underlies resistance to auranofin. Analysis of coding sequences with 100% SNVs frequency and at least 50-fold coverage identified a variety of ~5 non-synonymous SNVs in the exons of each line. Selected SNVs observed in the WGS were validated by PCR amplification and Sanger sequencing (primers in Table S1 ). We chose to specifically investigate the Aur^R line 8 to consider which SNV was most likely to underlie auranofin resistance. Line 8 has 12 SNVs that introduce amino acid substitutions to diverse protein coding genes ( Table 1 ). We used the community annotated ToxoDB database and BLAST analysis to investigate these loci. In particular, the availability of RNA-seq datasets permitted us to evaluate whether loci were expressed in the tachyzoite stage. In order to assess the significance of expressed loci, we used BLAST analysis to look for protein homologs and significant motifs and collected RNA-seq values and CRISPR screen scores for available loci. This information led us to focus on 2 of the loci (TGGT1_316330 and TGGT1_294640) as most likely resistance conferring. These loci encode a superoxide dismutase (SOD2) and the large subunit (R1) of ribonucleotide reductase (RNR). Both enzymes have strong fitness conferring effects in the deposited CRISPR screen data and well-conserved metabolic roles.

RNR removes the 2’-hydroxyl group of the ribose ring of nucleoside diphosphates to form deoxyribonucleotides from ribonucleotides which are used in DNA synthesis while SOD inactivates damaging superoxide ( O 2 − ) radicals by converting them into molecular oxygen and hydrogen peroxide. Since both TgRNR and TgSOD2 are conserved proteins, we threaded the amino acid sequence onto homologous crystal structures to locate the position of the amino acid substitutions. TgSOD2 is a Fe-type SOD and was previously demonstrated to localize to the tachyzoite mitochondrion (Brydges and Carruthers, 2003; Pino et al., 2007). SODs form dimers with shared coordination of two iron (FeIII) cofactors. The L201P mutation is located near to the key iron-coordinating residues: H111, E249 and H160 ( Figure 4A ) and is located in an α-helix. Substitution of a proline at this site likely causes the helix to bend, perhaps altering SOD enzymatic properties. TgRNR aligns well with human ribonucleotide reductase 1 (61.7% identity and 78.2% similarity). However, there is a 59 amino acid N-terminal extension to the Toxoplasma protein, with the conserved sequence beginning with M60 ( Figure 4B ). This leader does not represent mis-annotation of the N-terminus of the protein: there are MS peptides mapping to this sequence, especially in samples that are from monomethylarginine and phosphopeptide-enriched datasets in ToxoDB. This is particularly interesting as arginine methylation is a modification associated with proteins that have roles in transcriptional regulation, RNA metabolism and DNA repair. RNR use free-radical chemistry to convert ribonucleotides into deoxyribonucleotides. Significantly, ribose reduction requires generation of a free radical and RNR requires electrons donated from thioredoxin. Moreover, superoxide has been shown to inactivate RNR and yeast lacking either cytoplasmic or mitochondrial SODs have increased susceptibility to oxidative stress and RNR inactivation. The L166Q mutation is in a solvent exposed loop ( Figure 4C ), that is poorly conserved and not known to contribute to enzyme interactions.

The anti-parasitic effect of auranofin appears to stem from its molecule of gold, which is known to induce the accumulation of ROS in parasites (Debnath et al., 2012; Adeyemi et al., 2017). Since we hypothesized that a single mutation (SNV) suffices to generate resistance to auranofin, we analyzed the effect of the TgSOD2 L201P mutation on auranofin sensitivity. SOD2 is a bona fide enzymatic ROS scavenger and SOD2 is the sole mitochondrial superoxide dismutase in tachyzoites. To analyze the role of mutant SOD2 in auranofin resistance, we generated a T. gondii clone with a knock-in YFP tag for SOD2 that localizes to the tachyzoite mitochondrion as confirmed by co-localization with the mitochondrial marker F1B ATPase ( Figure 5A ). Subsequently, we generated two T. gondii SOD2.L201P lines: one lacking the YFP tag and one tagged with a C-terminal YFP fusion ( Figure S1 ). Both lines were corroborated by Sanger sequencing. In the absence of Auranofin, both L201P lines grew less well than the TgSOD2-YFP wild-type line but were more fit than Aur^R line 8 ( Figure 5B ). Previous attempts to create SOD1 and SOD2 gene deletions failed (Odberg-Ferragut et al., 2000) and SOD2 is a fitness-conferring locus in a genome-wide CRISPR/Cas screen (Sidik et al., 2016). It is noteworthy that the L201P mutation is associated with reduced fitness in the absence of superoxide stress (control conditions). In the presence of auranofin, only line 8 grows well, indicating that the single L201P mutation to TgSOD2 does not confer resistance. This may indicate that this substitution does not influence superoxide metabolism at all or that auranofin resistance also requires the mutant TgRNR enzyme. It is unlikely that these enzymes directly interact, as RNR enzymes are cytosolic while TgSOD2 is in the mitochondrial compartment. However, the SOD2.L201P T. gondii line accumulated ROS similarly to wild-type parasites, consistent with its auranofin-sensitive growth ( Figure 5C ).